AGP, PCI-E, PCI? Differences?

[Technofuzz]

dont forget that those celerons had less L1 cache too....64KB, vs pentium 4's 128KB....

 

[Technofuzz]

i think the p3's had 128KB L1 cache too, correct me if i'm wrong

 

[RadeonProVega]

I am looking at cpu-z for my celeron d and it reads L1 data cache 16kb, and L2 cache 2 at 256kb.
or am i missing something?

 

[Mussels]

I am looking at cpu-z for my celeron d and it reads L1 data cache 16kb, and L2 cache 2 at 256kb.
or am i missing something?

you're missing a lot of cache, because you have a celeron

 

[KainXS]

dont forget that those celerons had less L1 cache too....64KB, vs pentium 4's 128KB....

no, there is no pentium 4's with 128kb l1, not even 64

 

[RadeonProVega]

you're missing a lot of cache, because you have a celeron

Here is what it reads

My Celeron D

But compared to my P3, this rig is light years faster then my P3, so whatever cache is , it doesn't seem to be a problem.

My P3

http://img142.imageshack.us/img142/3671/20081217090726qj1.jpg
Look in the upper right

 

[Mussels]

cache clocks up with the CPU.

512KB at 2GHz is gunna be faster than 1MB at 1GHz, if you get my example.

and you're comparing different sockets with different chipsets and therefore far different overall system performance.

 

[KainXS]

does it matter, the most L1 cache I have seen a p4 with is like 28kb and that was the last pentium 4 made, your celeron d as 16KB just like all the other prescott celeron d's and pentium 4's but it has a quarter the l2 and is not quad pumped.

 

[Technofuzz]

Kain is right, lol i got the numbers with the new and old mixed up

 

[Technofuzz]

and you're comparing different sockets with different chipsets and therefore far different overall system performance.

OH REALLY? ever taken a 1GHz p3 laptop apart, identical CPU socket to P4....

 

[eidairaman1]

OH REALLY? ever taken a 1GHz p3 laptop apart, identical CPU socket to P4....

dont get me wrong
Desktop P4 and P3 are totally different socket wise, but i do recall that when the Pentium 4M Launched it was on a totally different Socket from the Desktop Part, and P3 Mobility was i believe SKT 370.

 

[Technofuzz]

newer P3 mobile CPUs were infact same as P4, trust me i know hardware...
and there were converters for 478 desktop mobos to accept the 479 socket p4-m

 

[Yukikaze]

newer P3 mobile CPUs were infact same as P4

Eh, what ?

 

[Technofuzz]

the newer models (2001 mobile P3s) have the same CPU socket as the mobile P4...

 

[Yukikaze]

the newer models (2001 mobile P3s) have the same mobile socket as the mobile P4...

Same socket isn't exactly: "newer P3 mobile CPUs were infact same as P4"...

The original statement was: "you're comparing different sockets with different chipsets and therefore far different overall system performance" and you nitpicked on the part of the sentence which matters the least. They might have the same sockets, but the CPUs themselves are widely different and difficult to directly compare.

 

[Technofuzz]

True....

But P4's were not alot faster than P3s, was just making a point that the P3's were remarkably quick and could hold their own against P4s; and towards that end they even upgraded the socket type for the mobile procs...

 

[Mussels]

True....

But P4's were not alot faster than P3s, was just making a point that the P3's were remarkably quick and could hold their own against P4s; and towards that end they even upgraded the socket type for the mobile procs...

so? P3 on SDram and P4 on DDR (or rambus) ram, while the P4 may have been slower in some cases, that doesnt mean that the memory subsystems werent a hell of a lot faster.

My point was in direct response to someone else: a more modern CPU with less cache can outperform an older one with more cache due to architectural differences between the CPU, the motherboard chipset, and hte overall system. Happy?

 

[eidairaman1]

you forget transistor density comes into play aswell, compare the fastest Athlon XP to the Slowest Athlon 64 and there is that much difference.

 

[Technofuzz]

A. some p4 boards used SDram....
B. P3s were faster then p4 (423 socket) in all cases
C. P3s were faster then p4 (478 socket) in some cases (depending on which proc's you compared)
D. ALL Celerons were terrible and SLOW, newer P3's meaning 1GHz+ could definitely hold their own against them, why intel killed the 478 celerons, ill never know....
E. DDR isn't a lot faster than SDram, due to case latency....

 

[RadeonProVega]

Quick question:

You guys said that pcie is much faster then PCI, obviously alright. So have any of you had a pci card and then tried the same card but in PCIe , how much performance difference do you see or for those who haven't done so, how much performance increase do you think you would see?

Example: The 8500GT series.

 

[eidairaman1]

you will run out of bandwidth on the PCI bus vs AGP, at higher resolution gaming but who really runs those cards on anything bigger than 1600x1200?

Also there were P3 Mobos that utilized Rambus and DDR.

 

[crazy pyro]

http://forums.techpowerup.com/showthread.php?t=88791&highlight=HD4350
Yukikaze benchmarked X16 being cut down to X1, X1 provides a similar bandwidth to PCI (I think so anyway, if it's different it'll be greater).

 

[farlex85]

Quick question:

You guys said that pcie is much faster then PCI, obviously alright. So have any of you had a pci card and then tried the same card but in PCIe , how much performance difference do you see or for those who haven't done so, how much performance increase do you think you would see?

Example: The 8500GT series.

Stop worrying about performance. You have an 8500gt, it's not going to likely make any noticeable difference what bus it is. You said you don't need the best, so why do you care about a frame or two? You want more performance, get a better card. Sheesh :shadedshu

 

[Yukikaze]

Quick question:

You guys said that pcie is much faster then PCI, obviously alright. So have any of you had a pci card and then tried the same card but in PCIe , how much performance difference do you see or for those who haven't done so, how much performance increase do you think you would see?

Example: The 8500GT series.

I don't have a direct comparison vs. a 8500GT on PCI-E, but:

My HD4350, sitting in a PCI-E x16 slot results in superior performance over my 8500GT on PCI by more than two-fold in 3DMark06 and Crysis. Hit the link in my sig to check the benchmarks, but the HD4350 reaches similar framerates in Crysis 1680x1050 at low as the PCI8500GT does at 1024x768 and 1680x1050 is 2.25 times the pixels on the screen !

The 9400GT is pretty much a 8500GT and the HD4350 is around 50% more powerful than the 9400GT. Neglecting any difference between the 8500GT and 9400GT and giving the 8500GT the benefit of the doubt means that the PCI bus cripples the 8500GT by almost two-fold.

Even on a PCI-E x1 slot (Gen1, not Gen2) the HD4350 absolutely annihilates the PCI8500GT.

PCI is a severe bottleneck even to weak cards, and that's why things like the PCI9500GT/PCI8600GT make even less sense.

 

[Yukikaze]

Stop worrying about performance. You have an 8500gt, it's not going to likely make any noticeable difference what bus it is. You said you don't need the best, so why do you care about a frame or two? You want more performance, get a better card. Sheesh :shadedshu

Actually, the difference is more like two-fold, rather than 2 more fps. Like I showed in my previous post: The HD4350 on PCI-E x16 beats the 8500GT on PCI in Crysis framerate when it is running at 1680x1050 and the PCI card is running 1024x768.